Re: Error: Need to obtain conf in GMPI_OPTS, and GMPI_CONF variables

[email protected] said on Fri, 21 Mar 2003 10:26 +0000:
> After upgrading our cluster we are getting the above error
> on _some_ of our nodes, even though they are virtualy identical.
> 
> RedHat 7.3 x86 (2.4.18-19.7.xsmp)
> gm-1.6.3
> mpich-1.2.4..8a
> OpenPBS_2_3_16
> maui-3.2.5
> 
> Innitally we didn't think it was a problem as our MPI jobs were running fine,
> however one or two users now think it may be causing issues.
> 
> I can't seem to figure out what they are supposed to do.
> Where/how are these variables supposed to be set ?
> 
> I would be grateful if someone could set me straight.

Probably you see that message on only some of your user codes, actually.
The message comes from libmpich.a.  This is what happens when you try to
run a code compiled against an old mpich library using a new mpiexec.
"Old mpich" means older than 1.2.4..8 which was released in Aug 2002 to
be precise.

What happened was that the Myricom completely changed the way they do
process startup in a way that is incompatible with earlier versions.
There does not seem to be a good way for mpiexec to poke at the
user's executable to guess which version of libmpich they're using.

You can tell users to recompile with your new mpich, or you can grab
mpiexec-0.69 and use that to launch these old-mpich jobs.  Search for
"completely incompatible" in the "Download" section of
    http://www.osc.edu/~pw/mpiexec/
to find the link to this older mpiexec.

		-- Pete